Types of maintenance and which to use for each case

The types of maintenance are listed in every manual, and almost none of them explain the one thing that actually matters: which one to use for which piece of equipment. Because they’re not a ladder where corrective is primitive and predictive is advanced. They’re different tools, and using the wrong one costs money in both directions.

The five types

Corrective. You intervene once the equipment has already failed.

Preventive. You intervene on a set schedule, regardless of the equipment’s condition.

Predictive, or condition-based. You intervene when a measurement indicates degradation.

Autonomous maintenance. The basic care the operator does themselves: cleaning, inspection, lubrication, small adjustments.

Improvement. Modifying the equipment or the procedure so the failure doesn’t happen again.

Which to use, depending on what drives the wear

This table settles almost every decision:

Equipment situation	Strategy
Failure with no real consequence, cheap or redundant	Corrective
Degrades over time, whether in use or not	Time-based preventive
Wears out with use	Counter-based preventive
Gives warning before failing and is critical	Predictive
Used by someone every day	Autonomous maintenance, in addition
Always fails for the same reason	Improvement
Set by a regulation	Preventive, at whatever frequency it requires

Almost every piece of equipment carries more than one at once. A compressor can have an annual time-based check, a filter change by operating hours, vibration monitoring by condition, and a daily operator inspection round.

Corrective isn’t the enemy

Worth saying, because the opposite gets repeated so often. For cheap, redundant equipment, or equipment whose failure has no consequence, corrective is the right strategy: preventively inspecting a light bulb costs more than just replacing it when it burns out.

The problem isn’t having corrective maintenance: it’s when that’s all there is, and when it happens it isn’t recorded. This is covered in corrective maintenance.

Preventive has two triggers

This is the distinction most often overlooked. Preventive maintenance can run by time — days and months — or by counter: the asset accumulates its hours, cycles or kilometers against a limit and a warning percentage, and when a reading crosses the threshold, the order is generated automatically.

Using only a calendar for equipment with very uneven usage is the most common cause of ending up with both overmaintained machines and machines that break down between checks at the same time.

Predictive doesn’t require sensors

The most expensive misunderstanding in the sector. A checklist with a minimum and maximum value turns a round into a time series: a reading outside range gets recorded as an anomaly right then, and with several readings you can see a trend.

One person with a thermometer and a portable vibration meter can do real predictive maintenance on critical equipment, at almost no cost. Continuous sensing speeds that up; it doesn’t invent it. More detail in predictive maintenance.

The reality in most companies

Here’s the part the manuals don’t tell you. The actual mix usually looks like this:

A lot of urgent corrective work, because it’s what drives the day.

Preventive maintenance on paper, which falls apart the moment a breakdown comes in, because a task that only lives on a calendar doesn’t protest when it isn’t done.

No predictive at all, because it’s perceived as expensive and technological.

Autonomous maintenance nonexistent or poorly framed, understood as pushing work onto production.

Improvement, which pays off the most, is the first thing to disappear when there’s no time.

And an added distortion: many plans have tasks that aren’t really maintenance — errands, admin, escorting a supplier — mixed in, inflating the executed plan percentage and making the preventive-to-corrective ratio look better than it actually is.

How to tell if your mix is unbalanced

With four figures from the reports:

Ratio of preventive to corrective hours. The most honest indicator, because it takes months to move.

Annual plan executed. If it’s far below target, the plan is bigger than the team can actually do, and that’s not fixed by pushing harder.

Anomalies detected. If it’s near zero, the checks are looking in the wrong place. If a piece of equipment has gone several cycles with none, it’s probably overmaintained.

Accumulated cost per asset. The figure that tells you which equipment deserves a different strategy.

How to fix it

Reassign strategy by family, not by individual asset. One afternoon’s conversation with whoever knows the assets: for each family, what drives its wear and what strategy it needs.

Cut back where there’s excess before adding where there’s a shortfall. Counterintuitive, and almost always the first move: freeing up hours from overmaintained equipment is what funds the rest.

Close the loop on anomalies. What gets detected has to turn into an issue with an owner and a date, or an explicit decision to do nothing. If it’s logged and nothing happens, people stop logging it.

Separate out what isn’t maintenance by type, so the metrics tell the truth.

Autonomous maintenance and improvement, which almost nobody uses

They deserve their own paragraph, because they’re the two with the best return per euro invested and the two least applied.

Autonomous maintenance isn’t cleaning in disguise, nor is it pushing work onto production: it’s the equipment operator taking on basic care — cleaning, visual inspection, lubrication, checking parameters — because they’re the only one who looks at it every day. A two-minute daily inspection catches things a six-monthly check can’t. This is covered in autonomous maintenance.

Improvement means modifying the equipment or the procedure so the failure doesn’t happen again, and it’s the only strategy that permanently reduces work instead of just reorganizing it. Its requirement is the data: without a failure history with its cause, “improving” stays an opinion.

What decides all of the above

The criticality of each asset, which is a field in its record in asset management: what stops production or service if it fails, what carries a risk to people, what carries a regulatory obligation.

Without that classification, the discussion over which strategy to apply repeats case by case. With it, it’s decided once, per family.

If you want to review what mix you’re currently running, you can request a demo on your own equipment.